Visualization and analysis of RNA-Seq assembly graphs.

Fahmi W Nazarie,Kim M Summers,Tom C Freeman,Tim Angus,Harpreet K Saini,Geoffrey J Faulkner,Sz-Hau Chen,Stijn Van Dongen,Karsten Klein,Anton J Enright,Mark W Barnett,Barbara Shih,Mick Watson

doi:10.1093/nar/gkz599

Abstract

RNA-Seq is a powerful transcriptome profiling technology enabling transcript discovery and quantification. Whilst most commonly used for gene-level quantification, the data can be used for the analysis of transcript isoforms. However, when the underlying transcript assemblies are complex, current visualization approaches can be limiting, with splicing events a challenge to interpret. Here, we report on the development of a graph-based visualization method as a complementary approach to understanding transcript diversity from short-read RNA-Seq data. Following the mapping of reads to a reference genome, a read-to-read comparison is performed on all reads mapping to a given gene, producing a weighted similarity matrix between reads. This is used to produce an RNA assembly graph, where nodes represent reads and edges similarity scores between them. The resulting graphs are visualized in 3D space to better appreciate their sometimes large and complex topology, with other information being overlaid on to nodes, e.g. transcript models. Here we demonstrate the utility of this approach, including the unusual structure of these graphs and how they can be used to identify issues in assembly, repetitive sequences within transcripts and splice variants. We believe this approach has the potential to significantly improve our understanding of transcript complexity.

Highlights

The advent of generation sequencing platforms enables new approaches to solving a variety of problems in medicine, agriculture, evolution and the environment
In order to explain the observed anomalies in the graph for this gene, we investigated the genomic origin of reads mapping to loop junctions using BLAST
Many tools and analysis pipelines already exist to process these data from the DNA sequencer, through mapping to a genome or de novo assembly, and summarize these data down to read counts per gene/transcript. These data are ready for differential gene expression or cluster-based analyses

Summary

Introduction

The advent of generation sequencing platforms enables new approaches to solving a variety of problems in medicine, agriculture, evolution and the environment. RNA-sequencing (RNA-Seq) based transcriptome analyses are used routinely as an alternative to microarrays for measuring transcript abundance, as well as offering the potential for gene and non-coding transcript discovery, splice variant and genome variance analyses [1,2]. Data are typically summarized by counting the number of sequencing reads that map to genomic features of interest, e.g. genes. These measures are used as the basis for determining the level of expression in a given sample and differential expression between samples. A large number of pipelines for the analysis of RNA-Seq data have been developed to go from the output of a sequencing machine, to sequence assembly, and on to the quantification of gene expression [3,4,5]. Many aspects of the analysis of these data remain computationally expensive and limiting, and tools are still under active development [6,7,8,9]

Methods

Results

Conclusion